Chap. 2: Datalog and SQL 2 IntelligentIntelligent Information Information Systems Systems WS 2018/19 DeductionDeductionDeduction ininin DatalogDatalogDatalog andandand SQLSQLSQL Chapter 2 © 201 8 Prof. Dr. Rainer Manthey Intelligent Information Systems 1 At the Core of Intelligent Databases: Managing Derived Data 2 • The essential topic of this lecture is to understand the importance of the idea of using the concept of derived data (and of the corresponding technology). derivedderived • We already learnt last week, that data - bases „containing “stored as well as derived data are called deductive databases in scientific terminology. • In order to manage derived data, we need special abilities of our DBMS that turn it into a deductive DBMS . • Managing derived data means • designing a deductive DB schema • incl. specifying derivation rules • evaluating queries over derived data storedstored • controlling consequences of changes of derived data © 201 8 Prof. Dr. Rainer Manthey Intelligent Information Systems 2 Stored or Derived (or both?) 2 • Storing data is the „normal “ way of keeping data (in some storage device). • Every data element can be stored (in principle). derivedderived • But there are data elements that do not (necessarily) have to be stored , but (possibly) can be computed from stored (or other) derived data elements instead. We call such pieces of data „derived “ data (elements). • It would be more precise to speak about „derivable “ data, because may be a derivable piece of data is actually never derived later on (or has never been derived up till now). We will use the term „derived “ data nevertheless. • Part of the derivable data can redundantly be stored (after having been derived first) in order storedstored to avoid costly rederivation. We call such data materialized (derived) data. © 201 8 Prof. Dr. Rainer Manthey Intelligent Information Systems 3 Another Topic of the Lecture: How to Derive Data? 2 derivedderived DBMS Inference Engine How do inference (derivation) engines look like? Should they be part of a DBMS or external tools? storedstored WeWe follow follow a a specific specificapproach approach in in IIS! IIS! © 201 8 Prof. Dr. Rainer Manthey Intelligent Information Systems 4 Learning by Doing 2 IntelligentIntelligent InformationInformation SystemsSystems WS 2018/19 2.2. Datalog Datalog and and SQL SQL 2.12.1 Datalog Datalog and and SQL: SQL: Learning Learning by by Doing Doing 2.22.2 Datalog: Datalog: Syntax Syntax and and (Basic) (Basic) Semantics Semantics 2.32.3 From From SQL SQL to to Datalog Datalog (and (and Back) Back) © 201 8 Prof. Dr. Rainer Manthey Intelligent Information Systems 5 Case Study: The British Royal Family 2 A case study on using derived data will give us a concrete start into using the techniques of Deductive Database technology: A Genealogical Database © 2017201 8 Prof. Dr. Rainer Manthey Intelligent Information Systems 6 The Royal Family Tree: Our Example Application Domain 2 Husband -Wife Parent -Child TheThe current current family family tree tree of of the the BritishBritish royal royal family family (starting (starting fromfrom the the Queen Queen and and her her husband) husband) © 2017201 8 Prof. Dr. Rainer Manthey Intelligent Information Systems 7 The Royal Family Tree: More Recent Updates 2 Husband -Wife Parent -Child SinceSince last last year, year, several several „ „updates“updates “ happened: happened: ••William William and and Kate Kate have have a a third third child, child, Louis Louis ••Harry Harry married married Meghan Meghan ••Eugenie Eugenie married married Jack Jack ••(to (to come: come: Meghan Meghan is is pregnant pregnant...) ...) © 2017201 8 Prof. Dr. Rainer Manthey Intelligent Information Systems 8 From Family Tree to Genealogical Database? 2 ?? HowHow to to „ „putput a a family family tree tree into into a a (relational) (relational) database database“?“? (i.e., how to design a basic genealogical database) © 2017201 8 Prof. Dr. Rainer Manthey Intelligent Information Systems 9 Conceptual Modeling of the Family Tree DB 2 • Traditional way of designing a relational DB: Start by conceptually modelling the resp. application domain • Corresponding ER diagram (Entity -Relationship approach): father husband parents child person marriage wife married mother title divorced name birth death sex • Basic technique for logical design : deriving a relational DB schema from an ER diagram • per entity type one table: Each attribute of the E -type turned into attribute of table • per relationship type on table, too: Each attribute of the R -type into attribute of corr. table, plus key attributes of of participating E -types (as foreign keys) © 2017201 8 Prof. Dr. Rainer Manthey Intelligent Information Systems 10 Logical Modeling of the Family Tree DB (1) 2 • The basic relational schema for this ER diagram looks as follows: • person (title, name , birth, death, sex) • parents ( father , mother , child ) • marriage ( husband , wife , married , divorced) • Roles in R -types are turned into attributes of the corr. R -table. father husband parents child person marriage wife married mother title divorced name birth death sex © 2017201 8 Prof. Dr. Rainer Manthey Intelligent Information Systems 11 Logical Modeling of the Family Tree DB (2) 2 • person (title, name , birth, death, sex) • parents (father, mother, child ) • marriage ( husband , wife , married , divorced) The initial design can be improved by integrating certain relati onship tables into entity tables in case of 1:N functionalities : father husband 1 N M parents child person marriage N wife 1 mother • person (title, name , birth, death, sex , father, mother) • marriage ( husband , wife , married , divorced) © 2017201 8 Prof. Dr. Rainer Manthey Intelligent Information Systems 12 Corresponding Genealogical Database 2 Person Key One possible (initial) relational format for family trees: just two tables! Title Marriage The entire family tree is „in “ these two tables. Disadvantage(?): Many empty cells ( „nulls “) © 201 8 Prof. Dr. Rainer Manthey Intelligent Information Systems 13 Just Stored Data in Our Genealogical DB – Up Till Now 2 • All the data representing what we know about the Royal family members (in the family tree) have to be stored in tables Title of our relational DB. • All the information in these two tables are „given “ data reported from the „real world “. • None of this could have been derived from other parts of the database. CanCan we we turn turn this this DB DB intointo a a deductive deductiveDB DB byby extending extending it it with with Marriage additionaladditional derived derived data data?? HowHowto to do do this, this, ... ... Person ...if...if we we can? can? © 2017201 8 Prof. Dr. Rainer Manthey Intelligent Information Systems 14 Genealogical Terminology: Potential for Deduction 2 Blood relation (or: consanguinity) only! Numbers indicate percentage of „common blood “ In this graph, relatives from the maternal side are given only – paternal relatives analogously. • Data about „existence “ (birth, death, sex, name etc.) cannot be derived , but have to be entered manually and stored in tables . The same applies to marriage and divorce. • All other data about family relationships, however, are derivable from these stored data, provided we can express them using suitable derivation rules . © 2017201 8 Prof. Dr. Rainer Manthey Intelligent Information Systems 15 Example: Deriving Uncle Data from the Stored Family Tree Data 2 e.g.: HowHow to to derive derive data data about about the the unclesunclesof of a a person? person? (Aunts could be treated analogously.) UncleUncle (from(from Latin: Latin: avunculus avunculus "little"little grandfather", grandfather", the the diminutive diminutive of of avus avus "grandfather")"grandfather") isis a a family family relationship relationship or or kinship, kinship, between between a a person person and and his his or or her her parent's parent's brother brother,, parent'sparent's brother-in-law brother-in-law or or parent's parent's cousin. cousin. (from: en/wikipedia) (Here we restrict ourselves to uncle in the narrow sense) © 2017201 8 Prof. Dr. Rainer Manthey Intelligent Information Systems 16 „Uncle “ as a Derived Table 2 Person UncleUncle Title ItIt would would be be nice nice to to havehave the the data data about about unclesuncles in in the the Royal Royal familyfamily in in a a table table of of itsits own own – –best best in in a a derivedderived table table!! (If(Ifpossible!) possible!) stored derived (Uncles of Charlotte and Mia still missing.) © 2017201 8 Prof. Dr. Rainer Manthey Intelligent Information Systems 17 Uncle Relationships in the Royal Family Tree? ?? AnAn uncleuncle ofofa a person person is is (one (one of) of) hishis (or (or her) her) parent's parent's brothers brothers.. © 201 8 Prof. Dr. Rainer Manthey Uncle Relationships in the Royal Family Tree! 2 !! AndAnd several several additional additionaluncle uncle relationshipsrelationships – –how how many? many? (Do Savannah and Isla have any uncles?) © 201 8 Prof. Dr. Rainer Manthey How to Derive a Particular Uncle Row? 2 UncleUncle Person HowHow to to derive derive the Uncle table Title the Uncle table fromfrom the the Person Person table?table? A derivation strate - gy a human would use, might serve as a suitable strategy of a deduction „engine “. (Is this some kind of „artificial intelligence “ ?) © 2017201 8 Prof. Dr. Rainer Manthey Intelligent Information Systems 20 Relational Views as Specifications of a Derivation Strategy 2 Derived tables in a relational DB are Title obtainable